Professional Documents
Culture Documents
and each weld pass fulfills the whole etched groove (no heat flux in
Introduction the z direction).
where:
Analytical Development T = temperature (oC)
In the one-dimensional model, the heat flux is considered to To = ambient temperature (oC)
occur only in the y direction, as shown in the coordinate system of θ = temperature difference (oC)
Fig. 1. The following assumptions are made: the heat source moves t = time (s)
at a sufficiently high speed (to neglect heat flux in the x direction), y = coordinate (m)
Q1 = thermal energy per unit area (J/m2)
α = thermal diffusivity of the material (m2/s)
ρ = density of the material(kg/m3)
Presented at COBEM 99 – 15th Brazilian Congress of Mechanical Engineering.
22-26 November 1999, São Paulo. SP. Brazil.
c = specific heat of the material (J/kgoC)
Paper accepted: July, 2002. Technical Editor: José Roberto de França Arruda.
∞
dQ y2
Q= ∫ q s ( y)dy (6) dθ( y, t ) = exp −
−∞ ρc 4παt 4αt
where:
Q = total power of the source (W) or
qs = power of the source per unit length (W/m)
d2
q s ( y′)dy′
dθ( y, t ) = exp − (11)
ρc 4παt 4αt
y
qmax
where d is the distance between the instantaneous source and a point
located on the y axis, that is,
d2 = (y-y’)2 (12)
C
Substituting Eqs. (10) and (12) in Eq. (11):
yb
3y 2 2
dθ( y, t ) =
3Q
exp − exp− ( y − y′) (13)
πy b ρc 4αt Erf ( 3 ) y2 4αt
b
(14)
q s ( y) = q max exp(−Ay 2 ) (7)
Solving the integral and rearranging the solution, one obtains:
where:
qmax = qs maximum value (W/m) 3Q
θ ( y, t ) =
A = coefficient of arc concentration (1/m) 2ρc π(12 αt + y 2b ) Erf ( 3 )
Coefficient A is determined considering a distance yb in Eq. (7),
exp − y2
+
y 2 y 2b 12 αt − yy b + y 2b
which corresponds to the distance from the origin to the location Erf + (15)
4αt 4α t (12 αt + y b )
2
2 α t (12 αt + y 2 )
where the power is reduced to five percent of its maximum value b
(Fig. 2). Thus,
exp − y 2 2 2
y yb 12 α t + yy b + y b
2
3 + Erf
A= (8) 4αt 4α t (12 αt + y 2b ) 2 α t (12 αt + y 2b )
y 2b
Equation (15) is the solution to the first weld pass, regarding the
When yb is large, qs(y) decreases slowly with y. Substituting Eq.
input of a heat source with Gaussian distribution. The solution to the
(8) in Eq. (7) and then in Eq. (6), and integrating this equation
second pass is obtained from the point heat source solution
between -yb and yb limits, one obtains:
(Suzuki,1996):
3Q d2
q max = (9) qs1 (y′)dy′ + qs2 (y′)dy′ exp− d2
π y b Erf ( 3 ) dθ(y, t) = exp −
ρc 4παt 4αt ρc 4πα(t − t p ) 4α(t − t p )
Equation (7) may then be written as: (16)
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright 2003 by ABCM July-September 2003, Vol. XXV, No. 3 / 303
Analytical Modeling of Multipass Welding Process with …
In Eq. (16), it may be observed that the t variable was displaced bead x width). Table 1 shows the physical properties used in the
by a value tp, which corresponds to the sum of the welding and simulation. It is known that the physical properties of the metal
waiting times to the beginning of the second pass. The use of indices change with temperature. However, this variation in the analytical
1 and 2 in the qs variable is for possible and sought variations of the models results in a non-linear equation, and it is not possible to
heat input between passes. The same steps applied to obtain Eq. (15) obtain the solution in closed form. Then, the physical properties are
are used to reach the solution for the second pass, and so on. usually taken at a specific temperature, for example, at half the
Analogously, the general solution to n passes, in terms of T, is given melting point of the material. In this work, they were calculated at
by: 800oC. The values refer to low carbon steels, but they can be used
for HSLA steel, as suggested by Hanz et al. (1989).
3 n Qi
T(y, t) = To + ∑ *
2ρc π i =1 12α[t − (i − 1)t ] + y2 Table 1. Physical properties of low carbon steels (Hanz et al,1989).
p b
k (J/msoC) ρc (J/m3oC) α (m2/s)
y2 y2 y2b
{exp− +
31.67 7.14x106 4.44x10-6
4α[t − (i − 1)t p 4α[t − (i − 1)t p{12α[t − (i − 1)t p ]}
Welding parameters. In order to fulfill the groove, in butt
12α[t − (i − 1)t p ] − yyb + y2b (17) welding, the usual practice is to increase the heat input from one
Erf + pass to the next. In the present simulation of a real case, the increase
2 α[t − (i − 1)t p ]{12α[t − (i − 1)t p + yb}
2
of heat input is obtained by increasing the welding current, the other
parameters in Eq. (18) remaining unaltered. However, the current
y 2 2 2
y yb
exp− + increase causes efficiency to decrease. Then, a different value of
4α[t − (i − 1)t p 4α[t − (i − 1)t p{12α[t − (i − 1)t p ]} efficiency must be used in each pass. The choices of these values
were based on the efficiency range for the Gas Metal Arc Welding
12α[t − (i − 1)t p ] + yyb + y2b (GMAW) process, which ranges from 66 to 85% (Svensson, 1994).
Erf } The welding parameters used in the simulation are in Table 2. The
2 α[t − (i − 1)t p ]{12α[t − (i − 1)t p + yb }
2
heat input (HI) values were determined by Eq. (18), multiplied by
the material thickness. The same HI values were used in the point
Equation (17) is the solution to the temperature distribution in and Gaussian heat source models. However, in the Gaussian heat
one-dimensional multipass welding processes, supplied by Gaussian source model, only ninety-five percent of HI was applied to the
heat sources. Far from the heat source, i.e., for distances where y is weld. In order to adjust such difference, the HI values in Table 2
of the same magnitude as yb, Eq. (16) is similar to the solution were multiplied by 1.05 for the Gaussian model.
obtained for the point heat source. However, near the FZ and HAZ
(y<<yb), the correction introduced by the distributed heat source Table 2. Welding parameters used three weld passes.
approach in Eq. (17) allows to better predicting the temperatures in
Pass I (A) V (V) v (m/s) η (%) HI (x106
these regions.
J/m)
1 186 26.4 80 0.79
Model Evaluation 2 235 26.2 0.005 75 0.92
3 301 25.4 70 1.07
A comparison between concentrated and distributed heat source
models was made through simulation of thermal cycles for three Parameter yb. In order to verify the capability of the proposed
weld passes. Equation (17) was used to calculate the temperatures model to reproduce the thermal cycles, some values were chosen for
near the fusion zone, when a Gaussian distributed heat sources is the y and yb variables. These choices took into account the heat
applied. The multipass model with concentrated heat source is given input used in the simulation, and also known values of yb from the
by Eq. (16). In this case, the variable qs(y) does not have a
literature, obtained via experimental determination. Table 3 shows
distribution, and it is calculated by:
the yb values obtained by Kou and Wang (1986), Zacharia et al.
(1989), and Wu (1992), as well as the heat input (HI) used in their
ηVI
qs = Q = (18) analyses. In the present work, the yb parameter was estimated based
vδ on the heat input values showed in Table 3. In Eq. (8), the A
coefficient was determined for the yb distance, where the power is
where: reduced to five percent of its maximum value. According to this
η = arc efficiency (%) equation, by increasing the heat input parameter yb also increases,
V = welding voltage (V) such that the area under the curve of Fig. 2 remain equal to ninety-
I = welding current (A) nine percent of its qmax value. Therefore, different values of yb for
v = welding speed (m/s) each pass were considered, since the heat input increased in the
δ = plate thickness (m) second and third passes. The yb values used in the simulation were
0.004; 0.0047 and 0.0054 m for the first, second and third pass,
The waiting time (time between passes) used corresponds to 60 respectively.
seconds, and the welding process was simulated during a total time
of 300 seconds. The ambient temperature is 25oC. The error Table 3. Heat input and yb values in the literature.
function in Eq. (17) was evaluated using a polynomial
approximation. The properties and parameters used in the simulation Author I V (V) V HI yb (m)
are described below. (A) (m/s) (x106J/m)
Material. The evaluation of the proposed model was made Kou and Wang (1986) 100 11 0.0055 0.20 0.003
considering butt welding of high strength low alloy steel (HSLA) Zacharia et al. (1989) 175 14 0.0034 0.72 0.003
plates, with dimensions 0.13 x 0.10 x 0.25 m (thickness x length Wu (1992) 200 20 0.010 0.40 0.002
1500
Results and Discussion
To compare the point and Gaussian heat source models, thermal
cycles were simulated at two different y locations, namely: y equal
to 0.001 m, appropriate for a position near the fusion zone (y < yb), 1000
point
and y equal to 0.003 m, a location at a distance of the same Gaussian
magnitude as the yb parameter.
T (oC)
Figure 3 shows the thermal cycles obtained after the point and
Gaussian heat source models, at y = 0.001 m. It can be observed that
the peak temperatures in each pass are higher for the point heat 500
source model than for the Gaussian heat source model. This occurs
due to the assumption that the heat input is instantaneously applied
over an infinitesimal volume cross-sectioned by the thickness-width
plane at the center of the workpiece. In the Gaussian heat source 0
model, it is assumed that the heat input is applied over the finite
0 100 200 300
volume, including the y coordinate at the extent of the yb parameter. t (s)
Therefore, the peak temperatures produced by the latter model is Figure 4.- Thermal cycles for the point and Gaussian heat source models,
expectedly more realistic. The previous determination of the peak at y = 0.003 m.
temperature to be reached at a specific location is interesting, since
it indicates fortuitous phase changes. The differences between the
temperatures simulated by the two models are described in the Table Conclusions
4. T1 refers to maximum temperature reached in the first pass; T2 in
the second pass, and so on. It is worth noticing that the correction in The main conclusions of this work are:
the proposed model affect only the peak temperature determination, - the closed form solution obtained allows to estimate the thermal
and no difference is seen in the cooling rate. cycles produced by multipass welding process, near the fusion and
heat affected zones;
- the distributed heat source in the proposed solution is an important
Table 4. Peak temperatures reached in each weld pass.
correction for the known model with point source, since this factor
Tpeak Point heat source Gaussian heat source allows to obtain temperature values more realistic near to the fusion
T1 2057.52 1632.28 zone;
T2 2535.16 1881.01 - the analytical solution derived from the point source model can be
T3 3044.18 2115.27 safely used to predict temperature fields away from the fusion zone
(FZ) and the heat affected zone(HAZ).
3500
Acknowledgement
3000
The authors want to acknowledge the financial support of
FAPESP - Fundação de Amparo à Pesquisa do Estado de São Paulo
2500 - to the research work reported in the present paper.
2000 point
References
T (oC)
Gaussian
1500 Goldak, J., Bibby, M., Chakravarti, A., 1984, A new finite element
model for welding heat sources, Metallurgical Transactions B, vol. 15B, pp.
299 - 305.
1000 Hanz, Z., Orozco, J., Indacochea, J. E., Chen, C. H., 1989, Resistance
spot welding : a heat transfer study, Welding Journal, vol. 68, n. 9, pp. 363s -
500 371s.
Kou, S., Wang, Y. H., 1986, Weld pool convection and its effect,
Welding Journal, vol. 65, n. 3, pp. 63s - 70s.
0
Pavelic, V., Tanbakuchi, R., Uyehara, O. A., Myers, P. S., 1969,
Experimental and computed temperature histories in Gas Tungsten Arc
0 100 200 300
t (s) Welding of thin plates, Welding Journal, vol. 48, n. 7, pp. 295s - 305s.
Rosenthal, D., 1941, Mathematical theory of heat distribution during
Figure 3. Thermal cycles for the point and Gaussian heat source models, welding and cutting, Welding Journal, vol. 20, n. 5, pp. 220s - 234s.
at y = 0.001 m. Suzuki, R. N., 1996, Solução Analítica para Distribuição de
Temperatura no Processo de Soldagem MIG com Múltiplos Passes, Tese
Mestrado, Universidade Estadual de Campinas, Campinas, São Paulo, Brasil.
In Figure 4, the thermal cycles were simulated using y = 0.003
Svensson, L. E., 1994, Arc welding process, in Control of
m. It is instrumental to show that, if y is close to yb, the Gaussian Microstructures and Properties in Steel Arc Welds, CRC Press.
heat source model is equivalent to the point heat source solution. Wu, C. S., 1992, Computer simulation of three-dimensional convection
The two models provide results that are practically the same. This in traveling MIG weld pools, Engineering Computations, vol.9, pp. 529 -
means that the point source model at distances far from the fusion 537.
zone can correctly predict the temperature fields. Zacharia, T., Eraslan, A. H., Aidun, D. K., David, S. A., 1989, Three-
dimensional transient model for arc welding process, Metallurgical
Transactions B, vol. 20B, pp. 645 - 659.
J. of the Braz. Soc. of Mech. Sci. & Eng. Copyright 2003 by ABCM July-September 2003, Vol. XXV, No. 3 / 305